pcb-mount air flow sensor

Use two thermistors, one heated by current flow thru it... refer to my BSEE thesis of 1962, where I used that technique to measure blood flow in the heart... unlike Obama's thesis, mine _is_ available in the Harvard Library (joint thesis MIT/Harvard Med School :-)

In later days, for GenRad portable equipment, I simply sensed the commutation current fall-to-zero glitch (DC, "electronic" fans) to flag a stalled or dragging fan.

In a more amusing situation, that sudden cessation of fan current at the commutation point was magnetically coupling to a _very_close_by_ CRT. Cured that by adding a feedback loop to make the fan "constant current". ...Jim Thompson

Fans with stall sensing are pretty common, these days. Why not use those?

Use three wire fans. That's what they're designed for. If the pulse output stopps, or drops below a certain rate then the fan is bad or blocked. Use one of SLoman's 555 as a missing pulse detector, if you can't just monitor it on an extra input on the MPU.

This will be on a couple of PCI Express crate controllers that plug into VME or VXI crates. We thought it would be cool if we reported crate power supply voltages, power supply noise, air temperature, and air flow. We don't have access to the fans.

We could use a couple of SOT23 transistors (something like 300 K/W in still air) running at the same currents but different dissipations, and look at the Vbe difference.

Yep. $4.99 at Fry's ;-) ...Jim Thompson

Or, if a DC fan, monitor the current. They're commutated by the rotation of the blades, and there is no internal filtering, so if the current stops wobbling then the fan has stopped spinning. Of course that does not detect the case of something blocking the intake or the pathological cases of someone cutting the blades off with a Dremel or extending the wires and moving the fan into another area.

Eg. this garage electric eye installation:

standard way is to compare the self heating in two temperature sensors, one exposed to the airflow, the other only to the air temperature. or just one selfheated sensor and compensate for temperature measured in some other way, like you suggest

you probably have one in you car, a platinum wire in the air intake held at a constant temperature, heating power and the air intake temperature gives airflow

well really air mass flow

-Lasse

Yep. Lift the hood of your vehicle and you'll find a MAF (mass air flow sensor). There are various methods. This shows most of them.

Basically, two hot wires in a bridge derangement. One in the air flow, and the other to act as a reference. Also used in weather stations for a no-moving-parts anemometer. They're very sensitive at low airspeeds, and

Or, just buy a commercial vane switch:

If you have a fair amount of CFM airflow to work with, a pitot tube, as used on aircraft, can be made to work. Add a pressure transducer or limit switch. I built one using a flex stainless pickup tube, acrylic pressure body, plastic ball, and emitter-optoisolator pair. If the ball hit bottom, it would block the light beam and trip the alarm. I was used to trigger on loss of air flow in a server rack plenum where the vendor did not want me "mounting" anything into their ducting. So, I found a hole, inserted the stainless pickup tube, and lived happily ever after.

If you're lazy, insert a 2nd fan in the airflow path. However, instead of using it for moving air, use it as a generator. When the generated voltage disappears, the air flow has stopped. Some butchery of the fan internal wiring will be required.

If the board is partly isolated from the air flow, you'll be measuring the thermal gradient across the board instead of the air flow. The sensors really need to be in a duct, or at least fairly close together.

As I mentioned in...

NNTP-Posting-Date: Tue, 31 Jul 2012 12:44:02 -0500 From: Jim Thompson Newsgroups: sci.electronics.design Subject: Re: pcb-mount air flow sensor Date: Tue, 31 Jul 2012 10:44:02 -0700 Message-ID:

Gads! ...Jim Thompson

At least some years ago, there were PTC thermistor devices that were made for this purpose. You applied a voltage with a series resistor, and the device went into thermal runaway if the airflow was too little. So, you had a large variation in voltage across the device due to a small change in heat removal. I think if you look for PTC thermistor you should be able to find such devices.

Jon

Are you measuring what you really care about? Is there some component that depends on air flow? Measure the temperature of that???

If you need an independent sensor, try this. You need a spare microcontroller pin, or one that can be shared by another button or sensor or indicator or...

Tie a thermistor to VCC. Cap to ground with the processor pin in the middle. Discharge the cap then measure the time for the pin to go high. Hold the cap at ground long enough to heat the thermistor and measure the time again. Time difference should depend on air flow.

the coolest no-moving-parts anemometer I've seen is basically three ultrasonic transceivers arranged in a triangle

you can get some that have more transceivers that measures air speed in three instead of two dimensions

-Lasse

I want to measure air flow across a PC board. It's in a card cage with a fan tray blowing air up through the boards, and we'd like to roughly measure flow and report it to our customers. Some surface-mount part would be ideal.

I like the thermistor idea, or is that too expensive? Maybe a thermocouple heated by current running through it? (Or two of them!)

George H.

Not too expensive. It would be nonlinear, and sensitive to ambient temperature as well as air flow.

A pair of back-to-back thermocouples would measure differential temperature (close enough) and one could be heated by a resistor or something. But I don't know of any surface-mount thermocouples, so we'd be doing hand assembly. It would need an opamp and an ADC to resolve microvolts.

I guess you could slam some voltage across a 1206 platinum RTD, then turn it off and measure temperature vs time as it cools off. Both theta and tau would depend on air flow. That sounds like fun, where engineering budgets constrain fun.

The two LM71s, one heated, isn't bad. All surface mount, SPI data, and it acquires ambient temp too. The trick would be to make sure the heater+sensor pair gets most of its cooling from air flow, and less from conduction into the copper planes of the PC board.

Rob has suggested two temp sensors with a heater between them. Air flow cools one sensor and heats the other. That does have a good zero-flow indication. It doesn't pick up ambient temp very well, so would need a third sensor for that.

tor

an

OK for Rob's idea turn off the heater to get the local temp.

You're not in any hurry to know the fan has stopper are you?

George H.

Two thermistors in series. One on a copper pour, the other not, measure voltage at the midpoint. Their resistance goes vaguely like exp(-0.03*T). With constant voltage drive, sensitivity will go up as exp(+0.03*T), because the resistor ratios will have the same dependence on delta-T, but the temperature rise will go as V**2/R.

With constant current, dissipation falls as T increases, so sensitivity will drop as well as voltage. So if you make the voltage go down a bit with increasing temperature, you can have constant bridge sensitivity.

Cheers

Phil Hobbs

tor

an

don't think you need to turn it off, if you know how much power you put in and what its resistance(temperature) is you get something that depends on air flow and air temperature which you could measure with another rtd with a much lower measurement current

so something like 2x current source, 2xRTD, 2xADC

there will be an offset for the power lost in the pcb so use thin tracks

-Lasse

Jeff Liebermann wrote in news: snipped-for-privacy@4ax.com:

I took apart a failed MAF from my Nissan Sentra,and there's considerable signal conditioning circuitry connected to the thin-film sensor. all on a ceramic substrate,imbedded in clear,flexible silicone goop.It has a bare IC chip (COB) about .25" square,and you can see the gold wires connecting it to the substrate pads.